Subject: Re: [HTCondor-users] correction: How to change machine availability for job execution?
From: Lee Gramling <leegramling@xxxxxxxxx> Date: 02/01/2016 10:37 AM
> I am just learning HTcondor and about to setup
a job queue with several
> linux machines. I have been reading the user manual and it appears
that
> for machines that can execute a job you put parameters into the
> condor_config for that machine.
> If I have several machines that can run jobs
but want to be able to make a
> machine unavailable perhaps thru a desktop gui and then make it available
> again later, do I just change the condor_config file locally?
A configuration best practices document would be helpful,
I think. For my site, I find that leaving condor_config essentially untouched
from the RPM distribution and doing everything in the config.d directory
is the best approach.
I number the files in rough categories by 100's, e.g.
000-099 is initialization stuff like the pool name, adminsitrators, etc.,
100's are for start _expression_ items, 200 is for machine characteristics,
300 is for job characteristics like submit_exprs, 400 is for hooks, and
so on.
You also will be well served to make the identical
configuration capable of being used on any system, by using the conditional
expressions for the config or defining machine classads based on attributes
like the hostname, IP address, or what have you, and then incorporate that
into your start _expression_. For example, IsDesktop may depend on the hostname
starting with "ws" (for "workstation") using the regexp()
ClassAd function, and then your start _expression_ would use that and the
time of day to determine whether to accept jobs: